The Correction Technologies For Off-line Handwritten Skewed Uyghur Text Lines In Document Images

As technology trends and paperless office needs, in order to improve efficiency and save storage space, the paper version documents will be phased out, in the past, a lot of important data are done by hand, now under the influence of the factors of the various techniques, handwritten text data needs to be converted by the OCR system for the electronic version of the material. This article will be process the off-line in Uyghur handwritten presences skew in text line. The research about off-line handwritten text line segmentation in text image is one of the hotspots in the most recent years in the field of pattern recognition, Particularly,with regard to off-line handwritten text Uyghur image processing. Before dividing lines of text, you need to process some situation below, something like varying writing style and habit, different skew angle and the phenomena of adhesions on the time of writing Uyghur. The purpose of this operating is to pave the way for text line segmentation, recognition and retrieval in following document image processing.In order to improve readability and recognition accuracy of Uyghur handwritten document images,It is very important for its skew correction. In this paper, a method to overcome the traditional Hough need to manually set different parameters, calculate the amount of heavy lifting and other shortcomings as a starting point, On the basis of the original heuristic learning methods, Using Otsu threshold algorithm to determine the core area of a line of text, and then use the least-squares fitting technique to get the lower baseline, to complete the picture correction by using a rotating algorithm. Therefore, processing the problems of text line skew exist in text images, it will lay a solid foundation for the subsequent text line segmentation. This technology was first used in the Uyghur handwritten text line skew correction, against the case of if the text image exits single text line and the overall text image skew, the result of the de-skew using the method proposed in this paper, the accuracy up to 98% even higher, in the end we got a very significant results.